Introduction to Gadfly

Gadfly is a library for plotting and visualization written in Julia. It is based largely on Hadley Wickhams's ggplot2 for R and Leland Wilkinson's book The Grammar of Graphics. Similar package in python is plotnine (https://realpython.com/ggplot-python/).

Some of the features are::

Renders publication quality graphics to SVG, PNG, Postscript, and PDF

Intuitive and consistent plotting interface

Works with Jupyter notebooks via IJulia out of the box

Tight integration with DataFrames.jl

Interactivity like panning, zooming, toggling powered by Snap.svg

Supports a large number of common plot types

Additional Recommended Resources:

Gadfly Documentation: http://gadflyjl.org/stable/

using CSV
using DataFrames
using Statistics
using FreqTables
using StatsBase
using DataFramesMeta
using Gadfly
using NamedArrays ##For named arrays

ENV["COLUMNS"] = 1000
ENV["LINES"] = 20

20

Case : Cars Data¶

This dataset is a slightly modified version of the dataset provided in the StatLib library. In line with the use by Ross Quinlan (1993) in predicting the attribute "mpg", 8 of the original instances were removed because they had unknown values for the "mpg" attribute. The original dataset is available in the file "auto-mpg.data-original".

"The data concerns city-cycle fuel consumption in miles per gallon, to be predicted in terms of 3 multivalued discrete and 5 continuous attributes." (Quinlan, 1993)

Attribute Information:

mpg: continuous
cylinders: multi-valued discrete
displacement: continuous
horsepower: continuous
weight: continuous
acceleration: continuous
model year: multi-valued discrete
origin: multi-valued discrete (1: American, 2: European 3: Japanese)
car name: string (unique for each instance)

More info: https://archive.ics.uci.edu/ml/datasets/Auto+MPG

Read the data¶

Read the data from the url source (fixed-width formatted lines)

homedir()

"/Users/Rahul"

pwd()

"/Users/Rahul/Documents/Rahul Office/IIMB/Concepts/Julia/ML_using_Julia/Julia_Code/Julia_Practice"

#autos_df = pd.read_csv('./data/autos_df.csv',
                       #index_col=['car_name'])
#autos_df.head()

autos_df = CSV.read("./data/autos_df.csv", DataFrame)

#autos_df.info()
eltypes(autos_df)

9-element Array{Type,1}:
 String
 Float64
 Int64
 Float64
 Union{Missing, Float64}
 Float64
 Float64
 Int64
 Int64

Typecast¶

convert converts from one Julia type to another, for things that “behave the same”. Float32 and Int64 are both * numbers, for the most part. Whereas a String and an Int64 are different things entirely.

autos_df[:,:origin] = convert.(Float64, autos_df[:,:origin])

398-element Array{Float64,1}:
 1.0
 1.0
 1.0
 1.0
 1.0
 1.0
 1.0
 1.0
 ⋮
 1.0
 1.0
 1.0
 2.0
 1.0
 1.0
 1.0

autos_df[:,:origin] = convert.(Int64, autos_df[:,:origin])

398-element Array{Int64,1}:
 1
 1
 1
 1
 1
 1
 1
 1
 ⋮
 1
 1
 1
 2
 1
 1
 1

string turns things into strings

autos_df[:, :s_origin] = string.(autos_df[:,:origin])

398-element Array{String,1}:
 "1"
 "1"
 "1"
 "1"
 "1"
 "1"
 "1"
 "1"
 ⋮
 "1"
 "1"
 "1"
 "2"
 "1"
 "1"
 "1"

parse turns a string into a Julia type. It only works on strings and not other types.

autos_df[:, :origin] = parse.(Int64, autos_df[:,:s_origin]);

autos_df[!, :origin] = parse.(Float64, autos_df[:,:s_origin])

398-element Array{Float64,1}:
 1.0
 1.0
 1.0
 1.0
 1.0
 1.0
 1.0
 1.0
 ⋮
 1.0
 1.0
 1.0
 2.0
 1.0
 1.0
 1.0

head(autos_df,2)

#head(autos_df,2)

Describe the features¶

Get the mean, median, mode etc

#autos_df.describe()

describe(autos_df)

horsepower seems to have missing values

#autos_df["horsepower"].isnull().values.any()
#autos_df[autos_df.horsepower.isnull()]

any(ismissing.(autos_df.horsepower))

true

#autos_df = autos_df.dropna()

autos_df = dropmissing(autos_df,:)

#autos_df[["mpg", "displacement","horsepower","weight","acceleration"]].describe()

describe(autos_df[:,["mpg", "displacement","horsepower","weight","acceleration"]])

What is the average miles per gallon for car with different origin

#autos_df.groupby('origin')['mpg'].mean()

combine(groupby(autos_df,:s_origin), :mpg .=> [mean])

Plot¶

Using Gadfly for different plots.

Distribution Plot¶

Histogram plot¶

Histogram is a tool to visualize one dimensional data which is continous in nature. Given a collection of single random variables:

Choose a interval (bins in which the entire dataset can be bucketed)

Count the data points within each bin (the y axis represents the frequency count)

plot function with Geom.histogram frokm gadfly package will render a histogram.

Plot the distribution of miles per gallon for cars with American origin and Japanese origin

autos_df[autos_df.s_origin .== "1",:]

#sn.distplot(autos_df[autos_df.origin == 1]['mpg'], label = 'American')
#plt.legend()

set_default_plot_size(24cm, 12cm)

origin_filter = autos_df[autos_df.s_origin .== "1",:]

hp1 = plot(origin_filter, 
            x="mpg", 
            Geom.histogram, 
    
            Guide.title("Cars of American Origin - 1"))

hp2 = plot(origin_filter, 
            x="mpg", 
            Geom.histogram(bincount=30), 
            
            Guide.title("Cars of American origin - 2"))

Plot the distribution of miles per gallon for cars from all origin

#sn.distplot(autos_df[autos_df.origin == 1]['mpg'],label = 'American', hist=True)
#sn.distplot(autos_df[autos_df.origin == 2]['mpg'], label = 'European',hist = True)
#sn.distplot(autos_df[autos_df.origin == 3]['mpg'], label = 'Japenese',hist = True)
#plt.legend()

##If plotting mpg for cars from american, european and japanese make.
hp3 = plot(autos_df, 
            x=:mpg,  
            color=:s_origin,
    
            Geom.histogram(position=:identity, bincount=40),
    
            Scale.color_discrete_manual("skyblue","red","green"),
            Theme(alphas=[0.5], discrete_highlight_color=identity),

            Guide.title("Cars with all origins"))

hstack(hp1, hp2, hp3)

Plot the histogram along with density

#Histogram with density plot
hp4 = plot(autos_df, 
            x=:mpg,  
            color=:s_origin,
    
            Geom.density(), 
            Geom.histogram(position=:identity, density = true, bincount=40),
    
            Guide.title("Histogram and Density Plot-1"),
    
            Scale.color_discrete_manual("skyblue","red","green"),
            Theme(alphas=[0.5], discrete_highlight_color=identity))

#The other way to write p4
hp5 = plot(autos_df, 
            x=:mpg,  
            color=:s_origin,
    
            layer(Geom.density(), Geom.histogram(position=:identity, density = true, bincount=40)),
    
            Guide.title("Histogram and Density Plot-1"),
    
            Scale.color_discrete_manual("skyblue","red","green"),
            Theme(alphas=[0.5], discrete_highlight_color=identity))

gridstack([hp4 hp5])

## Density plot using Geom.polygon 
hp6 = plot(autos_df, 
            x=:mpg,  
            color=:s_origin,
    
            Stat.density(),
            Geom.polygon(fill=true, preserve_order=true),
            Geom.histogram(position=:identity, density = true, bincount=40),
    
            Guide.title("Histogram and Density Plot-1"),
            Scale.color_discrete_manual("skyblue","red","green"),
            Theme(alphas=[0.5], discrete_highlight_color=identity))

##The other way to write p6
hp7 = plot(autos_df, 
            x=:mpg,  
            color=:s_origin,
    
            layer(Geom.density(), Geom.histogram(position=:identity, density = true, bincount=40)),
            
            ##Comment the above line nd uncomment the below one to see the change.
            #layer(Stat.density(), Geom.polygon(fill = true, preserve_order = true)),
            #layer(Geom.histogram(position=:identity, density = true, bincount=40)),
    
            Guide.title("Histogram and Density Plot-1"),
    
            Scale.color_discrete_manual("skyblue","red","green"),
            Theme(alphas=[0.5], discrete_highlight_color=identity))

gridstack([hp6 hp7])

#sn.distplot(autos_df[autos_df.origin == 3]['mpg'],label = 'Japenese',hist=False)
#plt.legend()

#sn.distplot(autos_df['mpg'],hist = False)

Exercise¶

Plot the distribution of horsepower for cars with American origin and Japanese origin

Kde Plot¶

Kernel density estimation(KDE) plot — plots a smooth curve shape of the distribution. It is a nonparametric estimation of density where inferences about the population is made from the finite data sample.

Parametric Data/Test: When the data is assumed to have been drawn from a particular distribution and some parametric test can be applied to it

Non-Parametric Data/Test: When we have no knowledge about the population and the underlying distribution

What is a Kernal?

Kernal: A kernel is a special type of probability density function (PDF) with the added property that it must be even. Thus, a kernel is a function with the following properties

non-negative

real-valued

even

its definite integral over its support set must equal to 1

Some common PDFs are kernels; they include the Uniform(-1,1) and standard normal distributions.

What is Kernal density estimation?

Kernel density estimation is a non-parametric method of estimating the probability density function (PDF) of a continuous random variable. It is non-parametric because it does not assume any underlying distribution for the variable. Essentially, at every datum, a kernel function is created with the datum at its centre – this ensures that the kernel is symmetric about the datum. The PDF is then estimated by adding all of these kernel functions and dividing by the number of data to ensure that it satisfies the 2 properties of a PDF:

Every possible value of the PDF (i.e. the function, f(x)), is non-negative.

The definite integral of the PDF over its support set equals to 1.

Steps in estimating kernal density:

Each observation is first replaced with a normal (Gaussian) curve centered at that value.

These curves are summed to compute the value of the density at each point in the support grid. The resulting curve is then normalized so that the area under it is equal to 1

More about KDE at:

https://chemicalstatistician.wordpress.com/2013/06/09/exploratory-data-analysis-kernel-density-estimation-in-r-on-ozone-pollution-data-in-new-york-and-ozonopolis/

Comparing mpg distributions of cars by different origins

#sn.distplot(autos_df[autos_df.origin == 1]['mpg'],hist=False, label = 'American')
#sn.distplot(autos_df[autos_df.origin == 2]['mpg'], hist = False, label = 'European')
#sn.distplot(autos_df[autos_df.origin == 3]['mpg'], hist = False, label = 'Japenese')

# Just the density plot
dp1 = plot(autos_df, 
            x=:mpg,  
            color=:s_origin,
    
            layer(Stat.density(), Geom.polygon(fill=true, preserve_order=true)),
    
            Scale.color_discrete_manual("skyblue","red","green"),
            Theme(alphas=[0.5], discrete_highlight_color=identity))

#Density plot with central 90% confidence limit.

dp2 = plot(autos_df, 
            x=:mpg,  
            color=:s_origin,
    
            #layer(Stat.density, Geom.polygon(fill=true, preserve_order=true), alpha=[0.4]),
            layer(Geom.density()),
            layer(Stat.quantile_bars(quantiles=[0.05, 0.95]), Geom.segment),
    
            Guide.title("1st Density plot with 90% CI"),
    
            Scale.color_discrete_manual("skyblue","red","green"),
            Theme(alphas=[0.2], discrete_highlight_color=identity))

##The other way
dp3 = plot(autos_df, 
            x=:mpg,  
            color=:s_origin,
    
            layer(Stat.density, Geom.polygon(fill=true, preserve_order=true) ), #alpha=[0.4]
            layer(Stat.quantile_bars(quantiles=[0.05, 0.95]), Geom.segment),
    
            Guide.title("2nd Density plot with 90% CI"),
    
            Scale.color_discrete_manual("skyblue","red","green"),
            Theme(alphas=[0.4], discrete_highlight_color=identity))

#hstack(dp1,dp2,dp3)
gridstack([dp1 dp2 dp3])

Bar plot¶

Plot average miles per gallon for different cylinder types using autos_df dataframe.

In Seaborm package (Python), barplot function has an estimator paramter which will anyways estimate the average value of a numeric feature for each categorical feature.

#sn.barplot(y = 's_origin',
#           x = 'cylinders',
#           data = autos_df,
#          )

If it is just a count of values w.r.t a variable:

plot(autos_df,
    x = "cylinders",
    Geom.bar,
    Theme(bar_spacing = 4mm)
    )

Here we have to compute it separate before using plot function.

To plot average miles per gallon for different cylinder types using DataFrames and Gadfly:

Use groupby method to group by cylinders and calulcate mean of mpg. Name this dataframe as mpg_cylinder_df.

call plot() from gadfly with Geom.bar to plot mpg_cylinder_df.

mpg_cylinder_df = combine(groupby(autos_df,[:cylinders]), :mpg .=> [mean])

bp1 = plot(mpg_cylinder_df, 
            x=:cylinders, 
            y=:mpg_mean,    
    
            Geom.bar(), #position=:dodge or :stack (but not needed for now.)
            
    
            Guide.title("1st Bar plot"),
    
    
            Theme( 
                  bar_spacing=4mm, 
                  key_position=:right),
    
            Coord.cartesian(xmin=2, xmax=9))

Draw the barplot for average miles per gallon grouped by cylinder and origin using DataFrames and Gadfly
1. Use groupby method to group by cylinders, origin and calulcate mean of mpg. Name this dataframe as mpg_cylinders_origin_df.
2. call plot() from gadfly with Geom.bar to plot mpg_cylinder_df.

#sn.barplot(x = 'cylinders',
#           y = 'mpg',
#           hue = 'origin',
#           data = autos_df)

mpg_cylinders_origin_df = combine(groupby(autos_df,[:cylinders,:s_origin]), :mpg .=> [mean])

mpg_cylinders_origin_df.label=string.(round.(Int, mpg_cylinders_origin_df.mpg_mean))

mpg_cylinders_origin_df

bp2 = plot(mpg_cylinders_origin_df, 
            x=:cylinders, 
            y=:mpg_mean,    
            color=:s_origin,
    
            
            label=:label,Geom.label(position=:centered),  Stat.dodge(position=:stack),
            Geom.bar(position=:stack),
           
            
    
            Guide.title("1st Bar plot"),
    
            Scale.color_discrete_manual("skyblue","red","green"),
    
            Theme( 
                  bar_spacing=4mm, 
                  key_position=:right),
    
            Coord.cartesian(xmin=2, xmax=9))

Exercise¶

Make the change in the above code to make the bar chart vertically dodged.

A horizontal dodged chart can be plotted as below:

bp3 = plot(mpg_cylinders_origin_df, 
            y=:cylinders, 
            x=:mpg_mean,    
            color=:s_origin,
    
            
            label=:label,Geom.label(position=:right),  Stat.dodge( axis=:y),
            Geom.bar(position=:dodge,orientation=:horizontal),
               
    
            Guide.title("2nd Bar plot"),
            Guide.yticks(orientation=:vertical), 
            Guide.ylabel("# Cylinders"),
            Guide.xlabel("Avg miles per gallon"),
    
            Scale.color_discrete_manual("skyblue","red","green"),
    
            Theme( 
                  bar_spacing=4mm, 
                  key_position=:right),
    
            Coord.cartesian(ymin=2, ymax=9, yflip=true))

Draw a barplot which shows the count of cars w.r.t it origin and number of cylinders the car has.

Use groupby method to group by cylinders, s_origin and and count the number of car. Name this dataframe as origin_cylinders_df.

call plot() from gadfly with Geom.bar to plot mpg_cylinder_df.

origin_cylinders_df = combine(groupby(autos_df,[:s_origin,:cylinders]), nrow .=> [:count])

bp4 = plot(origin_cylinders_df, 
            x=:cylinders, 
            y=:count,    
            color=:s_origin,
    
            
            layer(
            label=string.(origin_cylinders_df.count),
            Geom.label(position=:centered), 
            Stat.dodge(position=:dodge)),
    
            Geom.bar(position=:dodge),
           
            
    
            Guide.title("Count of cars with cylinders and origin"),
    
            Scale.color_discrete_manual("skyblue","red","green"),
    
            Theme( 
                  bar_spacing=4mm, 
                  key_position=:right),
    
            Coord.cartesian(xmin=2, xmax=9))

Heatmap¶

Can be used to visulize a square matrix, say a correlation matrix. To visulize a rectangular data, Gadfly has Geom.rectbin:

http://gadflyjl.org/stable/gallery/geometries/#[Geom.rect](@ref),-[Geom.rectbin](@ref)

We will proceed with visualizing a correlation matrix. Strictly speaking, Pearson’s correlation requires that each dataset be normally distributed, and not necessarily zero-mean.

Like other correlation coefficients, this one varies between -1 and +1 with 0 implying no correlation.

Correlations of -1 or +1 imply an exact linear relationship.

Positive correlations imply that as x increases, so does y.

Negative correlations imply that as x increases, y decreases.

Plot the correlation between 'mpg', 'displacement','horsepower','acceleration','weight'

autos_cor_df = autos_df[:,["mpg",
                        "displacement",
                        "horsepower",
                        "weight",
                        "acceleration"]]

head(autos_cor_df)

Below code using pandas will give correlation matrix in python

#auto_cor_df.corr()

cor(autos_df[:,"mpg"],autos_df[:,"horsepower"])

-0.7784267838977761

cor(Matrix(autos_cor_df))

5×5 Array{Float64,2}:
  1.0       -0.805127  -0.778427  -0.832244   0.423329
 -0.805127   1.0        0.897257   0.932994  -0.5438
 -0.778427   0.897257   1.0        0.864538  -0.689196
 -0.832244   0.932994   0.864538   1.0       -0.416839
  0.423329  -0.5438    -0.689196  -0.416839   1.0

cor_matrix = NamedArray( cor( Matrix(autos_cor_df) ) )

5×5 Named Array{Float64,2}
A ╲ B │         1          2          3          4          5
──────┼──────────────────────────────────────────────────────
1     │       1.0  -0.805127  -0.778427  -0.832244   0.423329
2     │ -0.805127        1.0   0.897257   0.932994    -0.5438
3     │ -0.778427   0.897257        1.0   0.864538  -0.689196
4     │ -0.832244   0.932994   0.864538        1.0  -0.416839
5     │  0.423329    -0.5438  -0.689196  -0.416839        1.0

cor_matrix = NamedArray( cor( Matrix(autos_cor_df)), 
            ([names(autos_cor_df);],[names(autos_cor_df);]), 
            ("Rows", "Cols"))

5×5 Named Array{Float64,2}
 Rows ╲ Cols │          mpg  displacement    horsepower        weight  acceleration
─────────────┼─────────────────────────────────────────────────────────────────────
mpg          │          1.0     -0.805127     -0.778427     -0.832244      0.423329
displacement │    -0.805127           1.0      0.897257      0.932994       -0.5438
horsepower   │    -0.778427      0.897257           1.0      0.864538     -0.689196
weight       │    -0.832244      0.932994      0.864538           1.0     -0.416839
acceleration │     0.423329       -0.5438     -0.689196     -0.416839           1.0

A developers way of doign the above. The below code block is not mine.

Source: https://discourse.julialang.org/t/first-impression-of-dataframes-jl/49753/5

struct NoPrint end; Base.show(::IO, ::NoPrint) = nothing
NamedArray([i > j ? cor(autos_cor_df[!, i], autos_cor_df[!, j]) : 
        NoPrint() 
        for i in 2:ncol(autos_cor_df), 
            j in 1:ncol(autos_cor_df)-1],
                  (names(autos_cor_df)[2:end], names(autos_cor_df)[1:end-1]))

4×4 Named Array{Any,2}
       A ╲ B │          mpg  displacement    horsepower        weight
─────────────┼───────────────────────────────────────────────────────
displacement │    -0.805127                                          
horsepower   │    -0.778427      0.897257                            
weight       │    -0.832244      0.932994      0.864538              
acceleration │     0.423329       -0.5438     -0.689196     -0.416839

End of developers code.

Proceeding with heatmap

Heatmap using seaborn in python.

#sn.heatmap(auto_clean_df.corr(),
#           annot = True,
#           cmap = sn.diverging_palette(250, 10, n = 25))

The basic heatmap using spy function from Gadfly

spy(cor_matrix)

cor_names= [names(cor_matrix,1);]  ##could have written (cor_matrix,2) as well

5-element Array{String,1}:
 "mpg"
 "displacement"
 "horsepower"
 "weight"
 "acceleration"

spy(cor_matrix, 
    Scale.y_discrete(labels = i->cor_names[i]), 
    Scale.x_discrete(labels = i->cor_names[i]),
    Scale.color_continuous(colormap=Scale.lab_gradient("red", "white", "green")))

scatter plot¶

Scatter plot is a cloud of points showing a joint distribution of two numerical variables where each point represents an observation from the dataset. Helps to understand the relationship between two numerical variables

Scatter plot mpg and weight of the cars

#sn.jointplot(x = 'mpg',
#             y = 'weight',
#            hue = 's_origin',
#             data = autos_df, 
#             kind = 'scatter'
#            )

Gadfly.push_theme(:dark) #default

plot(autos_df, x="mpg", y="weight", color = "s_origin", Geom.point)

The problem with scatter plot in over plotting. When dataset is huge, dots of the scatterplot tend to overlap, and graphic becomes unreadable and meanigless.

Hexabin plot¶

In one dimension straight line segments are the only possible shape for bin in a histogram. However for data in two dimensions bins can be more general shape (rectangular/Hexagon):

The obvious strategy is to choose a rectangular bin to build a histogram. Imagine the above scatter plot being filled with rectangular boxes (where the boxes represents the bins in horizontal and vertical direction). The count of values in each of the bins can be colored with gradient fill.

The hexagon tiling uses a hexagon shape for binning. The same scatterplot chart can be filled with hexagon shapes and the count of points falling in each hexagon can be used to fill the shape

Hex plot mpg and weight of the cars

Exercise¶

Using Geom.hexbin from Gadfly to draw the hexabin plot

This is how it is done in seaborn (Python):

#sn.jointplot(x = 'mpg',
#             y = 'weight',
#             data = autos_df, color = 'k',
#             kind = 'hex'
#            )

Boxplot¶

Box plot is a graphical representation of numerical data that can be used to understand the variability of the data and the existence of outliers.

A boxplot is a graph that gives you a good indication of how the values in the data are spread out.

To generate a box plot: Assume data as : 98, 77, 85, 88, 82, 83, 87, 67, 100, 63, 105

Arrange data in ascending order: 63, 67, 77, 82, 83, 85, 87, 88, 98, 100, 105 Calculate the median (middle value of the data, 85). This is Q2Calculate the median of the first half of the data, 77). This is Q1. *Calculate the median of the second half of the data, 98). This is Q3.

The box joins Q1 to Q3 (contains middle 50% of data).

IQR = Q3 - Q1 = 11

LIF = Q1 - 1.5*IQR = 60.5 ; UIF = Q3 + 1.5 IQR = 114.5

The point adjancent to LIF is 67 and point adjancent to UIF is 105.

The smallest observation greater than or equal to LIF builds lower whisker.

The largest observation less than or equal to UIF builds upper whisker.

Point outside the fences are outliers.

Intrepret boxplot:

If wide box and long whiskers, then maybe the data doesn’t cluster. If box is small and the whiskers are short, then probably your data does indeed cluster If box is small and the whiskers are long, then maybe the data clusters, but have some “outliers”

Plot the outliers in miles per gallon w.r.t. cylinders and origin.

Gadfly.push_theme(:default)

bp1 = plot(autos_df, x=:cylinders, y=:mpg, color=:s_origin,
    Geom.boxplot, Theme(boxplot_spacing=0.1mm),
    
    Guide.title("Boxplot of mpg with #cylinders"),
    Scale.color_discrete_manual("skyblue","red","green")
)

Concept of subgrid¶

What if we want to bring more than 4 dimensions of data in the same plot.

set_default_plot_size(24cm, 20cm)

subgrid1_df = combine(groupby(autos_df,[:s_origin,:cylinders]), :mpg .=> [mean])

gp1 = plot(subgrid1_df, 
            x=:s_origin, 
            y=:mpg_mean,    
            #xgroup=:cylinders, ##it can be any other categorical variable.
            ygroup=:cylinders,
            color=:s_origin, ## This can be some other categorical variable. Putting origin does not give any info.
    
            
            Geom.subplot_grid(
            layer(
            Geom.bar(position=:dodge)
            ),
        
            #To label the bars.
            layer(
            label=string.(round.(Int,subgrid1_df.mpg_mean)),
            Geom.label(position=:above)#, 
            #Stat.dodge(position=:dodge)
            )),
            
    
            Guide.title("Avg miles per gallon with cylinders and origin"),
    
            Scale.color_discrete_manual("skyblue","red","green"),
    
            Theme( 
                  bar_spacing=4mm, 
                  key_position=:right)
)

subgrid2_df = combine(groupby(autos_df,[:model_year,:s_origin,:cylinders]), :mpg .=> [mean])

tail(subgrid2_df)

gp2 = plot(subgrid2_df, 
            x=:model_year, 
            y=:mpg_mean,    
            #xgroup=:model_year, ##it can be any other categorical variable.
            ygroup=:cylinders,
            color=:s_origin, ## This can be some other categorical variable. Putting origin does not give any info.
    
            
            Geom.subplot_grid(
            layer(
            Geom.line()
            ),
        
            layer(
            label=string.(round.(Int,subgrid2_df.mpg_mean)),
            Geom.label(position=:dynamic, hide_overlaps=true)
            )),
            
            
    
            Guide.title("Avg miles per gallon with cylinders and origin"),
    
            Scale.color_discrete_manual("skyblue","red","green"),
    
    
            Scale.x_continuous(;minvalue=65, maxvalue=80),
    
            Theme(panel_stroke=colorant"black",
                  grid_line_width=0mm) ## we can use style() or theme to pass these parameters.
)

More on setting themes: http://gadflyjl.org/v0.7/man/themes.html#style-1

Exercise¶

Draw a boxplot using Geom.subplot_grid to split the boxplot with ygroup = s_origin. Refer Geom.subplot_grid used to plot barplot in above section.

Additional plots¶

Gadfly provides geometricall object to get fancy plots like

Violin plot: http://gadflyjl.org/stable/gallery/geometries/#[Geom.violin](@ref)
Beeswarm: http://gadflyjl.org/stable/gallery/geometries/#[Geom.beeswarm](@ref)
Contour plot: http://gadflyjl.org/stable/gallery/geometries/#[Geom.contour](@ref)
2dDensity plot: http://gadflyjl.org/stable/gallery/geometries/#[Geom.density2d](@ref)

and many more....

	car_name	mpg	cylinders	displacement	horsepower	weight	acceleration	model_year	origin
	String	Float64	Int64	Float64	Float64?	Float64	Float64	Int64	Int64
1	chevrolet chevelle malibu	18.0	8	307.0	130.0	3504.0	12.0	70	1
2	buick skylark 320	15.0	8	350.0	165.0	3693.0	11.5	70	1
3	plymouth satellite	18.0	8	318.0	150.0	3436.0	11.0	70	1
4	amc rebel sst	16.0	8	304.0	150.0	3433.0	12.0	70	1
5	ford torino	17.0	8	302.0	140.0	3449.0	10.5	70	1
6	ford galaxie 500	15.0	8	429.0	198.0	4341.0	10.0	70	1
7	chevrolet impala	14.0	8	454.0	220.0	4354.0	9.0	70	1
8	plymouth fury iii	14.0	8	440.0	215.0	4312.0	8.5	70	1
9	pontiac catalina	14.0	8	455.0	225.0	4425.0	10.0	70	1
10	amc ambassador dpl	15.0	8	390.0	190.0	3850.0	8.5	70	1
11	dodge challenger se	15.0	8	383.0	170.0	3563.0	10.0	70	1
12	plymouth 'cuda 340	14.0	8	340.0	160.0	3609.0	8.0	70	1
13	chevrolet monte carlo	15.0	8	400.0	150.0	3761.0	9.5	70	1
14	buick estate wagon (sw)	14.0	8	455.0	225.0	3086.0	10.0	70	1
15	toyota corona mark ii	24.0	4	113.0	95.0	2372.0	15.0	70	3
16	plymouth duster	22.0	6	198.0	95.0	2833.0	15.5	70	1
17	amc hornet	18.0	6	199.0	97.0	2774.0	15.5	70	1
18	ford maverick	21.0	6	200.0	85.0	2587.0	16.0	70	1
19	datsun pl510	27.0	4	97.0	88.0	2130.0	14.5	70	3
20	volkswagen 1131 deluxe sedan	26.0	4	97.0	46.0	1835.0	20.5	70	2
⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮

	variable	mean	min	median	max	nunique	nmissing	eltype
	Symbol	Union…	Any	Union…	Any	Union…	Union…	Type
1	car_name		amc ambassador brougham		vw rabbit custom	305		String
2	mpg	23.5146	9.0	23.0	46.6			Float64
3	cylinders	5.45477	3	4.0	8			Int64
4	displacement	193.426	68.0	148.5	455.0			Float64
5	horsepower	104.469	46.0	93.5	230.0		6	Union{Missing, Float64}
6	weight	2970.42	1613.0	2803.5	5140.0			Float64
7	acceleration	15.5681	8.0	15.5	24.8			Float64
8	model_year	76.0101	70	76.0	82			Int64
9	origin	1.57286	1.0	1.0	3.0			Float64
10	s_origin		1		3	3		String

	car_name	mpg	cylinders	displacement	horsepower	weight	acceleration	model_year	origin	s_origin
	String	Float64	Int64	Float64	Float64	Float64	Float64	Int64	Float64	String
1	chevrolet chevelle malibu	18.0	8	307.0	130.0	3504.0	12.0	70	1.0	1
2	buick skylark 320	15.0	8	350.0	165.0	3693.0	11.5	70	1.0	1
3	plymouth satellite	18.0	8	318.0	150.0	3436.0	11.0	70	1.0	1
4	amc rebel sst	16.0	8	304.0	150.0	3433.0	12.0	70	1.0	1
5	ford torino	17.0	8	302.0	140.0	3449.0	10.5	70	1.0	1
6	ford galaxie 500	15.0	8	429.0	198.0	4341.0	10.0	70	1.0	1
7	chevrolet impala	14.0	8	454.0	220.0	4354.0	9.0	70	1.0	1
8	plymouth fury iii	14.0	8	440.0	215.0	4312.0	8.5	70	1.0	1
9	pontiac catalina	14.0	8	455.0	225.0	4425.0	10.0	70	1.0	1
10	amc ambassador dpl	15.0	8	390.0	190.0	3850.0	8.5	70	1.0	1
11	dodge challenger se	15.0	8	383.0	170.0	3563.0	10.0	70	1.0	1
12	plymouth 'cuda 340	14.0	8	340.0	160.0	3609.0	8.0	70	1.0	1
13	chevrolet monte carlo	15.0	8	400.0	150.0	3761.0	9.5	70	1.0	1
14	buick estate wagon (sw)	14.0	8	455.0	225.0	3086.0	10.0	70	1.0	1
15	toyota corona mark ii	24.0	4	113.0	95.0	2372.0	15.0	70	3.0	3
16	plymouth duster	22.0	6	198.0	95.0	2833.0	15.5	70	1.0	1
17	amc hornet	18.0	6	199.0	97.0	2774.0	15.5	70	1.0	1
18	ford maverick	21.0	6	200.0	85.0	2587.0	16.0	70	1.0	1
19	datsun pl510	27.0	4	97.0	88.0	2130.0	14.5	70	3.0	3
20	volkswagen 1131 deluxe sedan	26.0	4	97.0	46.0	1835.0	20.5	70	2.0	2
⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮

	variable	mean	min	median	max	nunique	nmissing	eltype
	Symbol	Float64	Float64	Float64	Float64	Nothing	Nothing	DataType
1	mpg	23.4459	9.0	22.75	46.6			Float64
2	displacement	194.412	68.0	151.0	455.0			Float64
3	horsepower	104.469	46.0	93.5	230.0			Float64
4	weight	2977.58	1613.0	2803.5	5140.0			Float64
5	acceleration	15.5413	8.0	15.5	24.8			Float64

	car_name	mpg	cylinders	displacement	horsepower	weight	acceleration	model_year	origin	s_origin
	String	Float64	Int64	Float64	Float64	Float64	Float64	Int64	Float64	String
1	chevrolet chevelle malibu	18.0	8	307.0	130.0	3504.0	12.0	70	1.0	1
2	buick skylark 320	15.0	8	350.0	165.0	3693.0	11.5	70	1.0	1
3	plymouth satellite	18.0	8	318.0	150.0	3436.0	11.0	70	1.0	1
4	amc rebel sst	16.0	8	304.0	150.0	3433.0	12.0	70	1.0	1
5	ford torino	17.0	8	302.0	140.0	3449.0	10.5	70	1.0	1
6	ford galaxie 500	15.0	8	429.0	198.0	4341.0	10.0	70	1.0	1
7	chevrolet impala	14.0	8	454.0	220.0	4354.0	9.0	70	1.0	1
8	plymouth fury iii	14.0	8	440.0	215.0	4312.0	8.5	70	1.0	1
9	pontiac catalina	14.0	8	455.0	225.0	4425.0	10.0	70	1.0	1
10	amc ambassador dpl	15.0	8	390.0	190.0	3850.0	8.5	70	1.0	1
11	dodge challenger se	15.0	8	383.0	170.0	3563.0	10.0	70	1.0	1
12	plymouth 'cuda 340	14.0	8	340.0	160.0	3609.0	8.0	70	1.0	1
13	chevrolet monte carlo	15.0	8	400.0	150.0	3761.0	9.5	70	1.0	1
14	buick estate wagon (sw)	14.0	8	455.0	225.0	3086.0	10.0	70	1.0	1
15	plymouth duster	22.0	6	198.0	95.0	2833.0	15.5	70	1.0	1
16	amc hornet	18.0	6	199.0	97.0	2774.0	15.5	70	1.0	1
17	ford maverick	21.0	6	200.0	85.0	2587.0	16.0	70	1.0	1
18	amc gremlin	21.0	6	199.0	90.0	2648.0	15.0	70	1.0	1
19	ford f250	10.0	8	360.0	215.0	4615.0	14.0	70	1.0	1
20	chevy c20	10.0	8	307.0	200.0	4376.0	15.0	70	1.0	1
⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮

	s_origin	mpg_mean
	String	Float64
1	1	20.0335
2	3	30.4506
3	2	27.6029

	cylinders	mpg_mean
	Int64	Float64
1	8	14.9631
2	4	29.2839
3	6	19.9735
4	3	20.55
5	5	27.3667

	cylinders	s_origin	mpg_mean	label
	Int64	String	Float64	String
1	8	1	14.9631	15
2	4	3	31.5957	32
3	6	1	19.6452	20
4	4	2	28.1066	28
5	4	1	28.013	28
6	3	3	20.55	21
7	6	3	23.8833	24
8	6	2	20.1	20
9	5	2	27.3667	27

	s_origin	cylinders	mpg_mean
	String	Int64	Float64
1	1	8	14.9631
2	3	4	31.5957
3	1	6	19.6452
4	2	4	28.1066
5	1	4	28.013
6	3	3	20.55
7	3	6	23.8833
8	2	6	20.1
9	2	5	27.3667

Case : Cars Data¶

Read the data¶

Typecast¶

Describe the features¶

Plot¶

Distribution Plot¶

Histogram plot¶

Exercise¶

Kde Plot¶

Bar plot¶

Exercise¶

Heatmap¶

scatter plot¶

Hexabin plot¶

Exercise¶

Boxplot¶

Concept of subgrid¶

Exercise¶

Additional plots¶

Thank you¶